基于随机森林模型的需水预测模型及其应用

Water demand prediction model based on random forests model and its application

为解决需水预测模型精度问题,尝试基于随机森林模型的分类和回归功能构建需水预测模型。以苏州市需水量预测为研究实例,首先应用随机森林模型的分类功能将需水预测因子分类,经计算发现第一产业比例、人口、灌溉面积、万元产值用水量和国民经济生产总值为最重要的解释变量。在此基础上,用随机森林模型的回归功能对需水进行预测,同时采用相同的训练数据建立基于BP神经网络和RBF神经网络的需水预测模型,通过对比3个模型的预测结果,发现随机森林模型能有效预测需水量,且精度较高。

In order to improve the accuracy of a water demand prediction model, we attempted to use the classification and regression functions of the random forests model to construct a water demand prediction model. ＇Faking the water demand forecast in Suzhou City as a case study, we used the classification function to classify the water demand prediction factors, and found that the most significant explaining variables were the primary industrial structure, population, irrigation area, water use per 10 000 yuan, and GDP. On this basis, we used the regression fimction to predict the water demand, and used the same training data to construct the water demand prediction model based on the BP neural network and RBF neural network models. Through comparison of the prediction results of the three models, we drew the conclusions that the random forests model can effectively forecast the water demand, and it has higher precision than the other two models.